Generates the cluster-autoscaler-priority-expander ConfigMap expected by the Kubernetes Cluster Autoscaler's priority expander containing the Auto Scaling Group names rather than the names of the node groups.
As the Auto Scaling Groups don't change even when you update a Node Group configuration and perform a rolling update, this runs as a Job which has a timestamp in the name to run every Helm deploy.
This can be deployed with the Helm Chart in this repository. It must be deployed in the same namespace as cluster autoscaler. Currently a number of things including the input configmap are hardcoded in the chart.
The chart creates all the RBAC you need but it will need to have some AWS permissions to do the following:
"eks:DescribeNodegroup",
"eks:ListNodegroups"
These can be assigned via the EC2 instance profile, or assigned to the Pod via kiam, kube2iam or the EKS IAM Roles for service accounts feature.
Built as a temporary solution to solve: kubernetes/autoscaler#3871
Hopefully this issue will be fixed by AWS so this tool will no longer be required: aws/containers-roadmap#1304